BlockDex vs Shoogle: two ways to search across every shadcn registry

BlockDex vs Shoogle

2 compared on 6 criteria

Use Shoogle when you want the answer without leaving the terminal. It is the only one of the pair that registers as a shadcn namespace, so `npx shadcn@latest search @shoogle` searches from the CLI you are already running and the install follows in the same session — and the same index is served to any MCP client over streamable HTTP. Nothing about that route needs a browser, an account or a second tool. Use BlockDex when the question is which registries exist at all and whether the thing you found is free. Shoogle draws on the 292 community registries in shadcn's own directory; BlockDex goes looking, because "there is no directory of shadcn registries to read" that is complete — seeds, GitHub code search, GitHub topics and the ecosystem's awesome-list, with roughly two thirds of what that finds rejected before it is published. Two differences decide most cases. First, free versus paid: BlockDex settles it per item by fetching the exact URL the install would fetch, so free means the install has already been made once and unverified is a third answer rather than an optimistic guess — nothing else in this comparison answers it per item at all. Second, where the index lives: BlockDex publishes the whole corpus as JSON with no key and no account, so it can be queried, counted and built on, while Shoogle's registry endpoint answers a top hundred per query and publishes no total. Neither makes a claim about the quality of the code it points at, and neither should — they both point at somebody else's registry, and both label rather than rank.

compared2criteria6sources read5verified2026-09-06

Search every registry The API, no key

CriterionShoogleBlockDexWhat you get without payingEverything on the siteThere is no pricing page — https://shoogle.dev/pricing returned 404 when this row was checked — and no account gate on browsing, the directory or the MCP endpoint. The only commercial surface anywhere in the navigation is Sponsor, which is the same model SkillWorks runs and the opposite of a per-seat catalogue.All of it, including the APINo account, no key and no paid tier anywhere on the site, and the same is true of the index itself: "No key, no account, CORS open to everyone. Every screen on this site makes exactly one call, and these are the calls." The whole corpus is readable as JSON by anyone, which is the part none of the paid catalogues in this table offer at any price.Cheapest paid optionNot publishedNo plans, no tiers and no purchase surface. /pricing 404s and the MCP page's only call to action is the Discord. Read as sponsor-funded rather than free-forever: the site publishes no commitment either way, so this is what the page carries rather than what a visitor can infer.Not publishedThere is no pricing page and no paid tier. BlockDex does not sell components — it indexes the registries that do, including the paid ones, and says which is which.What it publishes292 registries; item count not published"S 292 registries" on the directory, over the same community set shadcn's own directory lists. The MCP page carries a UI-components counter and a registries counter, but both render as animated digit rollers rather than a number, so the page states no item total. The registry endpoint is bounded rather than complete in another way too: https://shoogle.dev/r/registry.json?q=button answers with pagination {"hasMore":false,"limit":100,"total":100} — a top-100 per query, not a corpus size.Every item in every public registry it can findItem-level rather than registry-level: the unit is the component, not the library it lives in. The live figures are deliberately not printed here — the crawl runs nightly, so any number in this table is wrong within a day. /api/stats is the current one. The method page publishes the coverage limit too: source is read for the first 150 items per registry, "a size decision, not a coverage one".How you install from itThrough the official shadcn CLI, as a namespace"Search with npx shadcn@latest search @shoogle — filter by type, limit, and offset without leaving your terminal." It is a shadcn registry namespace, so results come back already addressed to their SOURCE registry — the button row of that endpoint carries "addCommandArgument": "@animate-ui/primitives-buttons-button" — and the install runs against the origin registry rather than through Shoogle. The same index is also served as an MCP server over streamable HTTP for Claude, Cursor, Windsurf, VS Code and Cline. No other entry in this table is reachable from inside the CLI.Copy the shadcn CLI command off the rowEvery row carries the command that installs that item from its OWN registry — BlockDex is not in the install path and mirrors nobody's source. The command is verified rather than constructed: the crawler fetches "the exact URL npx shadcn@latest add would fetch", which is also how free and paid are decided, so a row that says free is a row whose install has already been made once.How something gets listedPulled from shadcn's own directory, plus submissions"Community registries pulled from the shadcn/ui directory, with sorting to make it easier to track newly added registries." So the inclusion decision is shadcn's, not Shoogle's, with a documented route for a registry that is not in it yet. Categories are assigned per registry — Design System, Marketing UI, AI Blocks and twenty more — and the sort is by when the registry was added.Nothing is listed; everything is crawled"Every number on this site is derived from something a registry actually served. There are no submissions, no editors and no way to buy a position." Registries are found four ways — seeds, GitHub code search for registry.json, GitHub topics, and the ecosystem's awesome-list mined for site links as well as repo links — and roughly two thirds of what that finds never reaches the index.What it says about the code's qualityNot publishedNothing on the directory or the MCP page claims review, curation, testing or any check of the code behind a listing. The one quality statement the site makes is about itself: every page carries an Alpha badge beside the wordmark.Whether it installs and whether it still exists"Crawled, not curated" — there is no taste judgement and no editorial ranking anywhere on the site. What is checked is mechanical and per item: whether the registry's own item endpoint serves the source (free), refuses it (paid), or was never reachable (unknown, and never rendered as free), and whether an item that was once indexed has since disappeared. An item that vanished keeps its page and says so, because that is the page someone arriving from a stale tutorial needs.

BlockDex is one of the 2 compared and publishes this table. "Not published" means that company does not state the figure on the page cited below. It is not a claim that no such figure exists: this table prints what was read, not what was inferred.

Can Shoogle be used from the shadcn CLI?

Yes, and it is the only entry in this comparison that can. Shoogle registers as a shadcn namespace, so `npx shadcn@latest search @shoogle` searches community registries from the terminal — its MCP page describes it as searching "without leaving your terminal", with filters for type, limit and offset. Results come back already addressed to the registry they live in, so the install runs against that registry rather than through Shoogle. The same index is also served as an MCP server over streamable HTTP for Claude, Cursor, Windsurf, VS Code and Cline. BlockDex has no CLI namespace; it is a website and an open API.

Which one covers more registries?

They are drawing on different pools. Shoogle's directory publishes "292 registries", and says where they come from: "Community registries pulled from the shadcn/ui directory" — so the inclusion decision is shadcn's. BlockDex assembles its own corpus because, as its method page puts it, "there is no directory of shadcn registries to read" that is complete: seeds, GitHub code search for registry.json from four angles, GitHub topics, and the ecosystem's awesome-list mined for site links as well as repository links. Roughly two thirds of what that finds is rejected before it is published. The current figures are on /api/stats rather than in this table, because the crawl runs nightly and any number printed here would be wrong within a day.

Does either one tell you whether a component is free?

BlockDex does, per item, and it is the only entry in this comparison that answers it at the item level at all. The crawler fetches "the exact URL npx shadcn@latest add would fetch": source comes back and the item is free, a 401 and it is paid, unreachable and it is unverified — a third answer, never rendered as free. Because that is the same request the install makes, a row marked free is a row whose install has already succeeded once. Shoogle publishes no free-or-paid signal on its directory or its MCP page.

Is there an API?

BlockDex publishes the whole index as JSON — "No key, no account, CORS open to everyone. Every screen on this site makes exactly one call, and these are the calls." Search, item detail, registry detail, the registry index and the stats row are all readable by anyone, which is what makes the figures on the site checkable rather than assertions. Shoogle's registry endpoint is reachable too, but it is shaped for the CLI rather than for reading the corpus: `/r/registry.json?q=button` answers with pagination `{"hasMore":false,"limit":100,"total":100}` — a top hundred for that query, not a count of what exists.

Do either of them vet the code they point at?

No, and neither claims to. Shoogle publishes nothing about review, curation or testing of a listing's code, and labels itself Alpha. BlockDex is explicit that it does not: "Crawled, not curated", with "no submissions, no editors and no way to buy a position". What it does check is mechanical — whether the item installs, and whether it still exists, with an item that has vanished from its registry keeping its page and saying so, because that is the page someone arriving from a stale tutorial needs. Both point at somebody else's registry, and both label rather than rank.

When you should pick one of these instead of BlockDex

BlockDex is not the right answer to every version of this question. Each case below is a real one.

Shoogle: You want the answer inside the tool you are already in rather than in a browser tab. Shoogle is the only one here that registers as a shadcn NAMESPACE, so `npx shadcn@latest search @shoogle` searches across community registries from the terminal and the install runs through the official CLI — no site, no copy-paste, and the same route works from any MCP client that speaks streamable HTTP. It is the shortest path from "is there a component for this" to the component being in the project. It also labels itself Alpha, and it publishes no figure for how many items that search covers.

What was checked, and when

Prices and limits move. Every source is printed here so a reader can check it on the day they read this, and a job re-fetches each of these URLs on a schedule and fails when a figure above no longer appears on the page it came from.

Shoogle, shoogle.dev/mcp: What you get without paying, Cheapest paid option, How you install from it. Read 2026-09-06.
Shoogle, shoogle.dev/directory: What it publishes, How something gets listed, What it says about the code's quality. Read 2026-09-06.
BlockDex, blockdex.thecompound.tech/api-reference: What you get without paying, Cheapest paid option. Read 2026-09-06.
BlockDex, blockdex.thecompound.tech/: What it publishes, What it says about the code's quality. Read 2026-09-06.
BlockDex, blockdex.thecompound.tech/method: How you install from it, How something gets listed. Read 2026-09-06.

Every cell above was confirmed against its own source on or since 2026-09-06. /blockdex-vs-shoogle